Queensland Spatial BioCondition Data Collection

The Spatial BioCondition 2021 version 2.0 dataset (SBC) was produced by the Queensland Herbarium and Biodiversity Science and the Remote Sensing Sciences business units in the Queensland Department of Environment, Science and Innovation. The pixel values in SBC dataset represent the predicted condition of vegetation for biodiversity in 2021. The range is 0-100, where lower values indicate poorer condition. No Data is represented by a value of 255. No Data include areas where: a. Regional Ecosystems with insufficient training and reference data to apply the framework b. Marine, intertidal, native grassland, sedgeland, forbland and predominantly unvegetated ecosystems defined in RE preclearing c. Urban, suburban, commercial, and industrial areas including intensive use lakes, estuaries, canals, dams and reservoirs as defined by Queensland Land Use Mapping dataset (https://qldspatial.information.qld.gov.au/catalogue/custom/detail.page?fid={BE30CE16-B1B9-48B1-BF21-DBE70597FA93}) Condition of vegetation for biodiversity may be influenced by agricultural practices, grazing land management, fire regimes and wildfire, urban development, incursion of invasive species, industrial logging, and mining. Queensland has a site-based vegetation condition assessment framework ‘BioCondition’, which assesses the relative capacity of an ecosystem to support the suite of species expected to occur in its relatively undisturbed (reference) state. This is measured using a suite of compositional, structural, and functional vegetation attributes which are compared against a reference. The greater the difference from the reference state the worse the condition. The reference state characteristics (the benchmark) are derived from a set of sites in the same vegetation community known to be in the best available condition. SBC moves the assessment of vegetation condition for biodiversity from a site-based approach to a predictive modelling approach that can be applied at the regional or state scale. It is based on the premise that the greater the difference (measured as distance in multi-dimensional remote sensing space) from the reference, the worse the condition. The model is developed using the remote sensing datasets as predictor variables and training sites with known RE and condition state as the response variable. The resulting model is applied to all vegetated areas with sufficient training and reference data to produce predictions of condition. The dataset comprises three bands. Band 2 is the predicted BioCondition score 0-100, with higher values representing better vegetation condition for biodiversity. Bands 1 and 3 show the lower and upper boundary of the 90% prediction interval, respectively. This prediction interval provides a likely range in which the true value of the prediction will be.

Data Creation
This dataset was created using the Spatial BioCondition modelling workflow. The model uses the following datasets: Sentinel 2 based green and bare fractional cover statistics (2019-2021); Landsat-derived fractional cover for the 2021 dry season; Sentinel 2 NDVI-derived phenological metrics; Landsat- and GEDI-derived canopy height (2019); Regional ecosystem pre-clearing dataset - version 13.0; and selected vegetation field survey data held in departmental databases
The final model output was clipped to the Brigalow Belt, Central Queensland Coast and Southeast Queensland biogeographic bioregion boundaries, and the following areas were masked: the pre-clearing extent of natural grasslands, sedgelands and forblands; predominantly unvegetated ecosystems; and regional ecosystems extra to the Brigalow Belt, Central Queensland Coast and Southeast Queensland bioregions defined by version 13.0 regional ecosystem mapping; built environments and infrastructure (urban, suburban, commercial and industrial areas including intensive use lakes, estuaries, canals, dams and reservoirs) defined by Queensland Land Use Mapping dataset (https://qldspatial.information.qld.gov.au/catalogue/custom/detail.page?fid={BE30CE16-B1B9-48B1-BF21-DBE70597FA93})
Masked areas, pixels without predictions, and pixels outside the bioregion are classified as No Data (DN = 255). Band 1 is the 5th percentile of the prediction interval, Band 2 is the predicted Spatial BioCondition score and Band 3 is the 95th percentile of the prediction interval